TASK	DATASET	MODEL	METRIC NAME	METRIC VALUE	GLOBAL RANK
Spectral Reconstruction	ARAD-1K	Restormer	PSNR	33.40	# 4
Spectral Reconstruction	ARAD-1K	Restormer	MRAE	0.1833	# 4
Spectral Reconstruction	ARAD-1K	Restormer	RMSE	0.0274	# 4
Grayscale Image Denoising	BSD68 sigma15	Restormer	PSNR	31.96	# 4
Image Denoising	DND	Restormer	PSNR (sRGB)	40.03	# 1
Image Denoising	DND	Restormer	SSIM (sRGB)	0.956	# 1
Image Defocus Deblurring	DPD	Restormer	Combined PSNR	25.98	# 4
Image Defocus Deblurring	DPD	Restormer	Combined SSIM	0.811	# 4
Image Defocus Deblurring	DPD	Restormer	LPIPS	0.178	# 3
Image Defocus Deblurring	DPD (Dual-view)	Restormer	PSNR	26.66	# 6
Image Defocus Deblurring	DPD (Dual-view)	Restormer	SSIM	0.833	# 6
Image Defocus Deblurring	DPD (Dual-view)	Restormer	LPIPS	0.155	# 4
Deblurring	GoPro	Restormer	PSNR	32.92	# 21
Deblurring	GoPro	Restormer	SSIM	0.961	# 19
Image Deblurring	GoPro	Restormer	PSNR	32.92	# 20
Image Deblurring	GoPro	Restormer	SSIM	0.961	# 19
Image Deblurring	GoPro	Restormer	Params (M)	26.13	# 12
Deblurring	HIDE (trained on GOPRO)	Restormer	PSNR (sRGB)	31.22	# 7
Deblurring	HIDE (trained on GOPRO)	Restormer	SSIM (sRGB)	0.942	# 8
Deblurring	HIDE (trained on GOPRO)	Restormer	Params (M)	26.13	# 8
Color Image Denoising	Kodak24 sigma50	Restormer	PSNR	30.01	# 2
Deblurring	MSU BASED	Restormer local	SSIM	0.94217	# 11
Deblurring	MSU BASED	Restormer local	PSNR	31.12341	# 8
Deblurring	MSU BASED	Restormer local	VMAF	65.25911	# 11
Deblurring	MSU BASED	Restormer local	LPIPS	0.08251	# 6
Deblurring	MSU BASED	Restormer local	ERQAv2.0	0.73875	# 11
Deblurring	MSU BASED	Restormer local	Subjective	0.1231	# 10
Deblurring	MSU BASED	Restormer	SSIM	0.94632	# 3
Deblurring	MSU BASED	Restormer	PSNR	31.76111	# 1
Deblurring	MSU BASED	Restormer	VMAF	66.3964	# 10
Deblurring	MSU BASED	Restormer	LPIPS	0.08239	# 4
Deblurring	MSU BASED	Restormer	ERQAv2.0	0.74776	# 3
Deblurring	MSU BASED	Restormer	Subjective	0.1175	# 11
Single Image Deraining	Rain100H	Restormer	PSNR	31.46	# 3
Single Image Deraining	Rain100H	Restormer	SSIM	0.904	# 4
Single Image Deraining	Rain100L	Restormer	PSNR	38.99	# 6
Single Image Deraining	Rain100L	Restormer	SSIM	0.978	# 8
Deblurring	RealBlur-J (trained on GoPro)	Restormer	PSNR (sRGB)	28.96	# 5
Deblurring	RealBlur-J (trained on GoPro)	Restormer	SSIM (sRGB)	0.879	# 5
Deblurring	RealBlur-R (trained on GoPro)	Restormer	PSNR (sRGB)	36.19	# 3
Deblurring	RealBlur-R (trained on GoPro)	Restormer	SSIM (sRGB)	0.957	# 2
Image Defocus Deblurring	RealDOF	Restormer	PSNR	25.08	# 2
Image Defocus Deblurring	RealDOF	Restormer	SSIM	0.769	# 2
Image Defocus Deblurring	RealDOF	Restormer	LPIPS	0.289	# 2
Deblurring	RSBlur	Restormer	Average PSNR	33.69	# 3
Image Denoising	SIDD	Restormer	PSNR (sRGB)	40.02	# 5
Image Denoising	SIDD	Restormer	SSIM (sRGB)	0.960	# 5
Single Image Deraining	Test100	Restormer	PSNR	32.00	# 1
Single Image Deraining	Test100	Restormer	SSIM	0.923	# 1
Single Image Deraining	Test1200	Restormer	PSNR	33.19	# 5
Single Image Deraining	Test1200	Restormer	SSIM	0.926	# 3
Single Image Deraining	Test2800	Restormer	PSNR	34.18	# 2
Single Image Deraining	Test2800	Restormer	SSIM	0.944	# 1
Color Image Denoising	urban100 sigma15	Restormer	Average PSNR	35.13	# 2
Grayscale Image Denoising	Urban100 sigma15	Restormer	PSNR	33.79	# 1
Grayscale Image Denoising	Urban100 sigma25	Restormer	PSNR	31.46	# 2
Color Image Denoising	Urban100 sigma50	Restormer	PSNR	30.02	# 3
Grayscale Image Denoising	Urban100 sigma50	Restormer	PSNR	28.29	# 3

Badge	Markdown
	`[![PWC](https://img.shields.io/endpoint.svg?url=https://paperswithcode.com/badge/restormer-efficient-transformer-for-high/image-denoising-on-dnd)](https://paperswithcode.com/sota/image-denoising-on-dnd?p=restormer-efficient-transformer-for-high)`
	`[![PWC](https://img.shields.io/endpoint.svg?url=https://paperswithcode.com/badge/restormer-efficient-transformer-for-high/single-image-deraining-on-test100)](https://paperswithcode.com/sota/single-image-deraining-on-test100?p=restormer-efficient-transformer-for-high)`
	`[![PWC](https://img.shields.io/endpoint.svg?url=https://paperswithcode.com/badge/restormer-efficient-transformer-for-high/grayscale-image-denoising-on-urban100-sigma15)](https://paperswithcode.com/sota/grayscale-image-denoising-on-urban100-sigma15?p=restormer-efficient-transformer-for-high)`
	`[![PWC](https://img.shields.io/endpoint.svg?url=https://paperswithcode.com/badge/restormer-efficient-transformer-for-high/color-image-denoising-on-kodak24-sigma50)](https://paperswithcode.com/sota/color-image-denoising-on-kodak24-sigma50?p=restormer-efficient-transformer-for-high)`
	`[![PWC](https://img.shields.io/endpoint.svg?url=https://paperswithcode.com/badge/restormer-efficient-transformer-for-high/image-defocus-deblurring-on-realdof)](https://paperswithcode.com/sota/image-defocus-deblurring-on-realdof?p=restormer-efficient-transformer-for-high)`
	`[![PWC](https://img.shields.io/endpoint.svg?url=https://paperswithcode.com/badge/restormer-efficient-transformer-for-high/single-image-deraining-on-test2800)](https://paperswithcode.com/sota/single-image-deraining-on-test2800?p=restormer-efficient-transformer-for-high)`
	`[![PWC](https://img.shields.io/endpoint.svg?url=https://paperswithcode.com/badge/restormer-efficient-transformer-for-high/color-image-denoising-on-urban100-sigma15-1)](https://paperswithcode.com/sota/color-image-denoising-on-urban100-sigma15-1?p=restormer-efficient-transformer-for-high)`
	`[![PWC](https://img.shields.io/endpoint.svg?url=https://paperswithcode.com/badge/restormer-efficient-transformer-for-high/grayscale-image-denoising-on-urban100-sigma25)](https://paperswithcode.com/sota/grayscale-image-denoising-on-urban100-sigma25?p=restormer-efficient-transformer-for-high)`
	`[![PWC](https://img.shields.io/endpoint.svg?url=https://paperswithcode.com/badge/restormer-efficient-transformer-for-high/single-image-deraining-on-rain100h)](https://paperswithcode.com/sota/single-image-deraining-on-rain100h?p=restormer-efficient-transformer-for-high)`
	`[![PWC](https://img.shields.io/endpoint.svg?url=https://paperswithcode.com/badge/restormer-efficient-transformer-for-high/deblurring-on-realblur-r-trained-on-gopro)](https://paperswithcode.com/sota/deblurring-on-realblur-r-trained-on-gopro?p=restormer-efficient-transformer-for-high)`
	`[![PWC](https://img.shields.io/endpoint.svg?url=https://paperswithcode.com/badge/restormer-efficient-transformer-for-high/deblurring-on-rsblur)](https://paperswithcode.com/sota/deblurring-on-rsblur?p=restormer-efficient-transformer-for-high)`
	`[![PWC](https://img.shields.io/endpoint.svg?url=https://paperswithcode.com/badge/restormer-efficient-transformer-for-high/color-image-denoising-on-urban100-sigma50)](https://paperswithcode.com/sota/color-image-denoising-on-urban100-sigma50?p=restormer-efficient-transformer-for-high)`
	`[![PWC](https://img.shields.io/endpoint.svg?url=https://paperswithcode.com/badge/restormer-efficient-transformer-for-high/grayscale-image-denoising-on-urban100-sigma50)](https://paperswithcode.com/sota/grayscale-image-denoising-on-urban100-sigma50?p=restormer-efficient-transformer-for-high)`
	`[![PWC](https://img.shields.io/endpoint.svg?url=https://paperswithcode.com/badge/restormer-efficient-transformer-for-high/spectral-reconstruction-on-arad-1k)](https://paperswithcode.com/sota/spectral-reconstruction-on-arad-1k?p=restormer-efficient-transformer-for-high)`
	`[![PWC](https://img.shields.io/endpoint.svg?url=https://paperswithcode.com/badge/restormer-efficient-transformer-for-high/grayscale-image-denoising-on-bsd68-sigma15)](https://paperswithcode.com/sota/grayscale-image-denoising-on-bsd68-sigma15?p=restormer-efficient-transformer-for-high)`
	`[![PWC](https://img.shields.io/endpoint.svg?url=https://paperswithcode.com/badge/restormer-efficient-transformer-for-high/image-defocus-deblurring-on-dpd)](https://paperswithcode.com/sota/image-defocus-deblurring-on-dpd?p=restormer-efficient-transformer-for-high)`
	`[![PWC](https://img.shields.io/endpoint.svg?url=https://paperswithcode.com/badge/restormer-efficient-transformer-for-high/deblurring-on-realblur-j-trained-on-gopro)](https://paperswithcode.com/sota/deblurring-on-realblur-j-trained-on-gopro?p=restormer-efficient-transformer-for-high)`
	`[![PWC](https://img.shields.io/endpoint.svg?url=https://paperswithcode.com/badge/restormer-efficient-transformer-for-high/image-denoising-on-sidd)](https://paperswithcode.com/sota/image-denoising-on-sidd?p=restormer-efficient-transformer-for-high)`
	`[![PWC](https://img.shields.io/endpoint.svg?url=https://paperswithcode.com/badge/restormer-efficient-transformer-for-high/single-image-deraining-on-test1200)](https://paperswithcode.com/sota/single-image-deraining-on-test1200?p=restormer-efficient-transformer-for-high)`
	`[![PWC](https://img.shields.io/endpoint.svg?url=https://paperswithcode.com/badge/restormer-efficient-transformer-for-high/image-defocus-deblurring-on-dpd-dual-view)](https://paperswithcode.com/sota/image-defocus-deblurring-on-dpd-dual-view?p=restormer-efficient-transformer-for-high)`
	`[![PWC](https://img.shields.io/endpoint.svg?url=https://paperswithcode.com/badge/restormer-efficient-transformer-for-high/single-image-deraining-on-rain100l)](https://paperswithcode.com/sota/single-image-deraining-on-rain100l?p=restormer-efficient-transformer-for-high)`
	`[![PWC](https://img.shields.io/endpoint.svg?url=https://paperswithcode.com/badge/restormer-efficient-transformer-for-high/deblurring-on-hide-trained-on-gopro)](https://paperswithcode.com/sota/deblurring-on-hide-trained-on-gopro?p=restormer-efficient-transformer-for-high)`
	`[![PWC](https://img.shields.io/endpoint.svg?url=https://paperswithcode.com/badge/restormer-efficient-transformer-for-high/deblurring-on-based)](https://paperswithcode.com/sota/deblurring-on-based?p=restormer-efficient-transformer-for-high)`
	`[![PWC](https://img.shields.io/endpoint.svg?url=https://paperswithcode.com/badge/restormer-efficient-transformer-for-high/image-deblurring-on-gopro)](https://paperswithcode.com/sota/image-deblurring-on-gopro?p=restormer-efficient-transformer-for-high)`
	`[![PWC](https://img.shields.io/endpoint.svg?url=https://paperswithcode.com/badge/restormer-efficient-transformer-for-high/deblurring-on-gopro)](https://paperswithcode.com/sota/deblurring-on-gopro?p=restormer-efficient-transformer-for-high)`

Restormer: Efficient Transformer for High-Resolution Image Restoration

CVPR 2022 · Syed Waqas Zamir, Aditya Arora, Salman Khan, Munawar Hayat, Fahad Shahbaz Khan, Ming-Hsuan Yang ·

Since convolutional neural networks (CNNs) perform well at learning generalizable image priors from large-scale data, these models have been extensively applied to image restoration and related tasks. Recently, another class of neural architectures, Transformers, have shown significant performance gains on natural language and high-level vision tasks. While the Transformer model mitigates the shortcomings of CNNs (i.e., limited receptive field and inadaptability to input content), its computational complexity grows quadratically with the spatial resolution, therefore making it infeasible to apply to most image restoration tasks involving high-resolution images. In this work, we propose an efficient Transformer model by making several key designs in the building blocks (multi-head attention and feed-forward network) such that it can capture long-range pixel interactions, while still remaining applicable to large images. Our model, named Restoration Transformer (Restormer), achieves state-of-the-art results on several image restoration tasks, including image deraining, single-image motion deblurring, defocus deblurring (single-image and dual-pixel data), and image denoising (Gaussian grayscale/color denoising, and real image denoising). The source code and pre-trained models are available at https://github.com/swz30/Restormer.

PDF Abstract CVPR 2022 PDF CVPR 2022 Abstract

Code

Add Remove Mark official

swz30/restormer official

↳ Quickstart in

Colab

Spaces

1,532

swz30/MPRNet

1,097

swz30/MIRNet

626

swz30/CycleISP

468

swz30/mirnetv2

↳ Quickstart in

Colab

367

See all 11 implementations

Tasks

Add Remove

Color Image Denoising

Deblurring

Denoising

Grayscale Image Denoising

Image Deblurring

Image Defocus Deblurring

Image Denoising

Image Restoration

Rain Removal

Single Image Deraining

Spectral Reconstruction

Vocal Bursts Intensity Prediction

Datasets

BSD

Urban100

GoPro

SIDD

Set12

McMaster

Synthetic Rain Datasets

Real Blur Dataset HIDE

DND DPD (Dual-view)

RSBlur ARAD-1K RealDOF

MSU BASED

Results from the Paper

Edit

Ranked #1 on Grayscale Image Denoising on Urban100 sigma15

Get a GitHub badge

Task	Dataset	Model	Metric Name	Metric Value	Global Rank	Benchmark
Spectral Reconstruction	ARAD-1K	Restormer	PSNR	33.40	# 4	Compare
			MRAE	0.1833	# 4	Compare
			RMSE	0.0274	# 4	Compare
Grayscale Image Denoising	BSD68 sigma15	Restormer	PSNR	31.96	# 4	Compare
Image Denoising	DND	Restormer	PSNR (sRGB)	40.03	# 1	Compare
Image Denoising	DND	Restormer	SSIM (sRGB)	0.956	# 1	Compare
Image Defocus Deblurring	DPD	Restormer	Combined PSNR	25.98	# 4	Compare
			Combined SSIM	0.811	# 4	Compare
			LPIPS	0.178	# 3	Compare
Image Defocus Deblurring	DPD (Dual-view)	Restormer	PSNR	26.66	# 6	Compare
			SSIM	0.833	# 6	Compare
			LPIPS	0.155	# 4	Compare
Deblurring	GoPro	Restormer	PSNR	32.92	# 21	Compare
Deblurring	GoPro	Restormer	SSIM	0.961	# 19	Compare
Image Deblurring	GoPro	Restormer	PSNR	32.92	# 20	Compare
			SSIM	0.961	# 19	Compare
			Params (M)	26.13	# 12	Compare
Deblurring	HIDE (trained on GOPRO)	Restormer	PSNR (sRGB)	31.22	# 7	Compare
			SSIM (sRGB)	0.942	# 8	Compare
			Params (M)	26.13	# 8	Compare
Color Image Denoising	Kodak24 sigma50	Restormer	PSNR	30.01	# 2	Compare
Deblurring	MSU BASED	Restormer local	SSIM	0.94217	# 11	Compare
			PSNR	31.12341	# 8	Compare
			VMAF	65.25911	# 11	Compare
			LPIPS	0.08251	# 6	Compare
			ERQAv2.0	0.73875	# 11	Compare
			Subjective	0.1231	# 10	Compare
Deblurring	MSU BASED	Restormer	SSIM	0.94632	# 3	Compare
			PSNR	31.76111	# 1	Compare
			VMAF	66.3964	# 10	Compare
			LPIPS	0.08239	# 4	Compare
			ERQAv2.0	0.74776	# 3	Compare
			Subjective	0.1175	# 11	Compare
Single Image Deraining	Rain100H	Restormer	PSNR	31.46	# 3	Compare
Single Image Deraining	Rain100H	Restormer	SSIM	0.904	# 4	Compare
Single Image Deraining	Rain100L	Restormer	PSNR	38.99	# 6	Compare
Single Image Deraining	Rain100L	Restormer	SSIM	0.978	# 8	Compare
Deblurring	RealBlur-J (trained on GoPro)	Restormer	PSNR (sRGB)	28.96	# 5	Compare
Deblurring	RealBlur-J (trained on GoPro)	Restormer	SSIM (sRGB)	0.879	# 5	Compare
Deblurring	RealBlur-R (trained on GoPro)	Restormer	PSNR (sRGB)	36.19	# 3	Compare
Deblurring	RealBlur-R (trained on GoPro)	Restormer	SSIM (sRGB)	0.957	# 2	Compare
Image Defocus Deblurring	RealDOF	Restormer	PSNR	25.08	# 2	Compare
			SSIM	0.769	# 2	Compare
			LPIPS	0.289	# 2	Compare
Deblurring	RSBlur	Restormer	Average PSNR	33.69	# 3	Compare
Image Denoising	SIDD	Restormer	PSNR (sRGB)	40.02	# 5	Compare
Image Denoising	SIDD	Restormer	SSIM (sRGB)	0.960	# 5	Compare
Single Image Deraining	Test100	Restormer	PSNR	32.00	# 1	Compare
Single Image Deraining	Test100	Restormer	SSIM	0.923	# 1	Compare
Single Image Deraining	Test1200	Restormer	PSNR	33.19	# 5	Compare
Single Image Deraining	Test1200	Restormer	SSIM	0.926	# 3	Compare
Single Image Deraining	Test2800	Restormer	PSNR	34.18	# 2	Compare
Single Image Deraining	Test2800	Restormer	SSIM	0.944	# 1	Compare
Color Image Denoising	urban100 sigma15	Restormer	Average PSNR	35.13	# 2	Compare
Grayscale Image Denoising	Urban100 sigma15	Restormer	PSNR	33.79	# 1	Compare
Grayscale Image Denoising	Urban100 sigma25	Restormer	PSNR	31.46	# 2	Compare
Color Image Denoising	Urban100 sigma50	Restormer	PSNR	30.02	# 3	Compare
Grayscale Image Denoising	Urban100 sigma50	Restormer	PSNR	28.29	# 3	Compare

Methods

Add Remove

Absolute Position Encodings • Adam • BPE • Dense Connections • Dropout • Label Smoothing • Layer Normalization • Linear Layer • Multi-Head Attention • Position-Wise Feed-Forward Layer • Residual Connection • Scaled Dot-Product Attention • Softmax • Transformer

Edit Social Preview

Restormer: Efficient Transformer for High-Resolution Image Restoration

Code Edit Add Remove Mark official

Tasks Edit Add Remove

Datasets Edit

Results from the Paper Edit

Methods Edit Add Remove

Code

Add Remove Mark official

Tasks

Add Remove

Datasets

Results from the Paper

Edit

Methods

Add Remove